Method for Proactive Comprehensive Geriatric Risk Screening

1. A method of performing proactive comprehensive geriatric risk screening comprising: receiving at a processing device, individual features data of a patient being assessed for multiple risk types; running, by the processing device, a multi-task predictive model trained to jointly predict multiple target risk types for said individual based on said individual features data and predict a set of risk associations by determining correlations between target risk types, said multi-task predictive model trained based on: data representing risks across multiple vulnerability domains, data representing features of multiple patients, and data representing complete or incomplete observations in risk targets and features of said multiple patients; and trained based on optimizing, at the processing device, a linkage regularization using the features data, the risks across multiple vulnerability domains data and said complete or incomplete observations data, said linkage regularization regulating said multi-task predictive model training, a selecting and ranking of said risk features, and a learning and selecting the set of risk associations, by linking a coefficient matrix relating target features and risk types used in said predictive model and a covariance matrix representing domain knowledge on risk associations; and calculating, by the processing device, a risk score for said jointly predicted multiple target risk types for said individual patient using said trained model and said linkage regularization optimizing; and outputting said patient risk score for each said multiple target risk types in each domain for said individual patient for display via a device providing a user interface; and providing, based on the predicted patient risk score for said multiple target risk types, a course of preventative treatment for the individual patient.

2. The method of claim 1 , wherein said optimizing said linkage regularization comprises: performing an iterative algorithm on said risk feature selection and ranking; applying a thresholding rule to update elements of the covariance matrix representing domain knowledge on risk associations used by the iterative algorithm for the risk feature selection and ranking; and leveraging said performing the iterative algorithm and said applying the thresholding rule.

3. The method of claim 2 , wherein said iterative algorithm comprises: running a smoothing proximal gradient algorithm.

4. The method of claim 1 , wherein the training further comprises: receiving one or more of expert opinion data, and domain knowledge on risk association data.

5. The method of claim 1 , further comprising: determining, by said processor, whether a score of a particular risk target for said individual patient is one of: a high-risk score or low risk score.

6. The method of claim 1 , wherein the individual features comprise at least one of electronic medical records, answer data from a questionnaire administered to said patient, genetics information, activity data, and diet tracking.

7. An apparatus for performing proactive comprehensive geriatric risk screening, the apparatus comprising: a memory storage device storing a program of instructions; a processor device receiving said program of instructions to configure said processor device to: receive individual features data of a patient being assessed for multiple risk types; run a multi-task predictive model trained to jointly predict multiple target risk types for said individual based on said individual features data and to predict a set of risk associations by determining correlations between target risk types, said multi-task predictive model trained based on: data representing risks across multiple vulnerability domains, data representing features of multiple patients, and data representing complete or incomplete observations in risk targets and features of said multiple patients; and trained based on optimizing linkage regularization using the features data, the received risks across multiple vulnerability domains data and said complete or incomplete observations data, said linkage regularization regulating said multi-task predictive model training, a selecting and ranking of the risk features, and a learning and selecting of the set of risk associations, said linkage regularization linking a coefficient matrix relating target features and risk types used in said predictive model and a covariance matrix representing domain knowledge on risk associations; and calculate a risk score for said jointly predicted multiple target risk types for said individual patient using said trained model and said linkage regularization optimizing; and output said patient risk score for each said multiple target risk types in each domain for said individual patient for display via a device providing a user interface; and provide, based on the predicted patient risk score for said multiple target risk types, a course of preventative treatment for the individual patient.

8. The apparatus of claim 7 , wherein the processor device is further configured to: perform an iterative algorithm on said risk feature selection and ranking; apply a thresholding rule to update elements of the covariance matrix representing domain knowledge on risk associations used by the iterative algorithm for the risk feature selection and ranking; and leveraging said perform the iterative algorithm and said apply the thresholding rule.

9. The apparatus of claim 8 , wherein said iterative algorithm comprises a smoothing proximal gradient algorithm.

10. The apparatus of claim 7 , wherein the processor device is further configured to: receive one or more of expert opinion data, and domain knowledge on risk association data.

11. The apparatus of claim 7 , wherein the processor device is further configured to determine, whether a score of a particular risk target for said individual patient is one of: a high-risk score or low risk score.

12. The apparatus of claim 7 , wherein the individual features comprise at least one of electronic medical records, answer data from a questionnaire administered to said patient, genetics, activity data, and diet tracking.

13. A non-transitory computer readable storage medium, tangible embodying a program of instructions executable by the computer for performing proactive comprehensive geriatric risk screening comprising: receiving individual features data of a patient being assessed for multiple risk types; running a multi-task predictive model trained to jointly predict multiple target risk types for said individual based on said individual features data and predict a set of risk associations by determining correlations between target risk types, said multi-task predictive model trained based on: data representing risks across multiple vulnerability domains, data representing features, and data representing complete or incomplete observations in risk targets and features of said multiple patients; and trained based on optimizing linkage regularization using the features data, the risks across multiple vulnerability domains data and said complete or incomplete observations data, said linkage regularization regulating said multi-task predictive model training, a selecting and ranking of the risk features, and a learning and selecting of the set of risk associations, said linkage regularization linking a coefficient matrix relating target features and risk types used in said predictive model and a covariance matrix representing domain knowledge on risk associations; and calculating a risk score for said jointly predicted multiple target risk types for said individual patient using said trained model and said linkage regularization optimizing; and outputting a patient risk score for each said multiple target risk types in each domain for said individual patient for display via a device providing a user interface; and providing, based on the predicted patient risk score for said multiple target risk types, a course of preventative treatment for the individual patient.

14. The non-transitory computer readable storage medium of claim 13 , wherein optimizing said linkage regularization comprises: performing an iterative algorithm on said feature selection and ranking; applying a thresholding rule to update elements of the covariance matrix representing domain knowledge on risk associations used by the iterative algorithm for the risk feature selection and ranking; and leveraging said performing the iterative algorithm and said applying the thresholding rule.

15. The non-transitory computer readable storage medium of claim 14 , wherein said iterative algorithm comprises a smoothing proximal gradient algorithm.

16. The non-transitory computer readable storage medium of claim 13 , wherein the training further comprises receiving one or more of expert opinion data, and domain knowledge on risk association data.

17. The non-transitory computer readable storage medium of claim 13 , further comprising: determining whether a score of a particular risk target for said individual patient is one of: a high-risk score or low risk score.